Radiomics: Extracting more Features using Endoscopic Imaging

1.

Radiomics: Extracting more
Features using Endoscopic Imaging

2.

Presentation summary
Conclusion
Descusion
Endoscopic images
Computer Aided detection systems
Topics
What is “Radiomics”?
Computed Tomography images

3.

Computer Aided detection systems

4.

What is Radiomics?
Radiomics refers to the conversion of images into mineable information
and also the analyze that information for decision support.
Histogram analysis
3.
4.
1. Mean
2. median
maximum intensity
minimum intensity
Texture Feature
1- Contrast
2- Correlation
3- Spectral
4- Homogeneity
Color Based feature
Color Histogram
Histogram Intersection
Color Histogram for K means
Color Correlogram
Chromaticity
Shape Based features
Perimeter of the Boundary
Diameter of Boundary
Eccentricity
Curvature
Topological Descriptors

5.

Computed Tomography images
CT images are overcome the matter of
superimposition of organs, bones, and another parts
of body in any depths, by taking many images of the
region of interest with variety angles.
CT IMAGING
Advantages
Limits
› detect mucosal tissues and nodules
› The applied radiations to the body in CT scans
› easier to make difference between the
bones of the back and front side
› CT Scan represents data at a particular point
of time

6.

Endoscopic images
Endoscopy is a technic to make medical
image through the endoscope and a
camera at the end of the scope connected
to a larger monitor.
Endoscopic
IMAGING
Advantages
Limits
› provides direct and clear field visualization of
the disease.
› endoscopy cannot make images from
inside of muscles
› endoscopy is a minimally invasive procedure
› depth perception is not possible with
endoscopes

7.

Discussion
In medical specialty, features of cancers detected from radiological data (e.g. CT
scans, endoscopy images and MRI) are often use to process detection, prediction,
and prognostic cancer in patient.
Histogram analysis
CT Images
Endoscopic Images
Texture Feature
HSV color histogram
(Hue, Saturation, Value)
Shape Based
features
RG (Red Green) Texture
Color Based feature
LBP texture
(Local Binary Pattern)

8.

Conclusion
According to the features extracted from the medical images reviewed in this
Presentation, endoscopic images are capable of extracting color features that
significantly improve the performance of the cancer detection system.
However, due to the limitations of the use of endoscopic imaging, which cannot
detect diseases that are primarily involve the submucosa, muscular, or serosal
layers of the intestine also, if suspected that the bowel is punctured, it is not a
good procedure, endoscopic images cannot be used to detect all types of
cancers. Thus, if endoscopic and CT scan images are available, the processing of
endoscopic images will increase the efficiency of cancer detection as compared
with CT scan.

9.

Ready to answer your questions
Faridoddin Shariaty
E-mail: [email protected]
Phone number: +79522770591